Back to skill

Security audit

xiaodu-movie-mode-official

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed XiaoDu movie-mode home automation orchestrator, with some expected caution around broad triggers and saved preferences.

Install this only if you want XiaoDu movie-mode phrases to adjust connected home devices such as lights, curtains, TV/projector, volume, and signal source. Review any saved movie preferences periodically, and avoid enabling it in shared spaces where a broad phrase like “I want to watch a movie” could unexpectedly change the room state.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad natural-language requests like '我要看电影' and '准备看电影了', which can occur in ordinary conversation and may invoke the skill without the user intending full home-automation changes. In this skill's context, activation can lead to lighting, curtains, display devices, and other room-state changes, so accidental invocation has real-world side effects beyond a harmless informational response.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The description does not clearly warn users that invoking the skill may automatically control home devices such as lights, curtains, TVs, projectors, and room environment settings. Because this is a physical-world automation skill, missing disclosure increases the chance of uninformed consent, surprise actuation, and unsafe or unwanted changes in the home.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The test cases explicitly require writing persistent user preferences such as default device, volume, and HDMI behavior into XIAODU_CONTEXT.md, but they do not mention any user notice, consent flow, retention rule, or way to review/delete that data. In a home-automation skill, these preferences can reveal household habits and cause silent long-term behavior changes, so undocumented persistence is a real privacy and transparency issue.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The activation phrases include broad, everyday utterances such as '我要看电影' and '帮我安排观影', which can be said casually or ambiguously and may trigger home-environment changes without sufficiently specific intent. In this skill context, unintended activation can cause device state changes like dimming lights, closing curtains, and powering TV/projector equipment, making the issue more significant than a harmless misclassification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The notes instruct the system to absorb user behavioral preferences and write them into XIAODU_CONTEXT.md, but provide no guidance for notice, consent, retention limits, or sensitivity filtering. In a household assistant context, these preferences can reveal routines and in-home habits, so silent persistence creates a privacy and profiling risk even if the collected data seems operationally useful.

Content

No source excerpt is available for this finding.